These 100% AI generated posts about taste have to be some elaborate performance art piece, I genuinely do not understand how someone could think it a good idea "write" this.
In either case, taste being all that's left doesn't really help at all since whatever FAANG or top startups do is by definition good taste. Same for fashion, art and so on with the top being suitably defined. Slim fit was all the rage until all the uncool people got in and that's what ended the trend and not any friction or utility.
It's also just patently false. Taste is not all that is left. Are software engineers really so full of hubris that they thought coding is all that there is to making a product?
> Are software engineers really so full of hubris that they thought coding is all that there is to making a product?
Frankly, this barely scratches the surface of the hubris of engineers. I personally think AI has been one of the best thing to happen to software engineering. Writing boilerplate or my millionth auth implementation was never why I fell in love with the craft in the first place.
We have computers, computers can automate things. Boilerplate is a sign of a bad language (not a bad engineer). But then we add infinite automation features to reduce boilerplate and we get C++, and C++ programmers decry the excessive number of features and want to delete them all and go back to writing boilerplate by hand.
I once wrote a pile of C++ templates that would let you wrap a C++ function as a JavaScript function, automatically converting all the arguments and the return type. Then I left the company. How did my successors find it? No idea. Maybe they pulled out all the automation complexity and replaced it with manually operated boilerplate.
This is true, but I was mostly referring to framework boilerplate (Spring, React, etc.) or plumbing boilerplate. If you've ever written code professionally, you know that most code that you write is just making your thing fit with someone else's thing.
If your consumer or your provider made bad engineering decisions, you have to absorb them. If your company mandates the use of a framework, you have to absorb that, and so on. So I think it's great that AI can write all the dumb shit I didn't come up with anyway.
> So let me be plain about it: Claude was not here. No LLM wrote this—not a sentence of it, nor was it outlined, drafted, reviewed, checked, etc. by one, and there is no prompt behind it either
Following form of the words I just read: these 100% animous filled HN threads about how everyone is bad and everything deserves mocker have to be some elaborate power game to undermine hope and human will, I genuinely do not understand how genuine human people could think it good to "comment" like this.
There's just such a ridiculous pile on of people with such enormous overwhelming negativity underneath this ocmment. This off-gassing is almost never engaged with the material. Now there's an excuse to write off any material we don't like instantly, to accuse it of fakeness. But I still do not see a realness enscribed upon people who so actively work to tear down & prevent thought & consideration.
To engage with the post: I do think the definition of fashion is interesting, and a real power dynamic to be aware of. But do think there's a lot of things that take off in the world because they are good and interesting and nice, especially in technology and products, that the world wants nice things, and will happily support you for doing that. And especially in tech, the technical decisions we start from and build underfoot steer us towards success in subtle compounding ways, that matter, and that taste is crucial, where fashion is not a factor.
Again, at a meta level: my distress & frustration at the top is less from seeing this comment, and far more from seeing everything down thread trying to out-do one another with how cynical and anti-meaning anti-trying they can be. But I think this post definitely started on an all too familiar note, blood in the HN waters that all too often gets us so sharky.
IMO giving "benefit of the doubt" to something beyond doubt is bad, straight up. If someone can't tell this article is AI then they need to go spend an hour a day reading known AI writing vs known human writing until they develop the skill; the voice of this author SCREAMS AI.
I'm a firm believer in not engaging with or promoting the content of misrepresented AI.
On the topic of taste, I think of this quote in Susan Sontag's "Notes On 'Camp'" often:
“For taste governs every free — as opposed to rote — human response. Nothing is more decisive. There is taste in people, visual taste, taste in emotion — and there is taste in acts, taste in morality. Intelligence, as well, is really a kind of taste: taste in ideas. (One of the facts to be reckoned with is that taste tends to develop very unevenly. It's rare that the same person has good visual taste and good taste in people and taste in ideas.)
Taste has no system and no proofs. But there is something like a logic of taste: the consistent sensibility which underlies and gives rise to a certain taste. A sensibility is almost, but not quite, ineffable. Any sensibility which can be crammed into the mold of a system, or handled with the rough tools of proof, is no longer a sensibility at all. It has hardened into an idea..."
This article resonated tremendously with me, speaking as someone who has been coding since the 1980s and had to develop my own taste the hard way, through lots and lots of mistakes. Whenever I see a software demo built by someone mostly using agents to do the work, I have to question whether it's any good on the inside - or whether any real intuition or judgment went into its design. On the flip side, maybe none of that matters anymore - as long as it works, who cares how it was built.
It resonated with me too, and even as a once fairly prolific coder that now uses agents to do all the coding, I feel like this is all relevant and does matter. The agent does whatever you ask and provides thin judgements based on what it thinks is your motivation and context, even feeding it the context it will often make bad generic choices. I think this is the part hard to optimize because it's not a verifiable reward, it's the feel of the software. I think Herdr is a great example, I've tried several dozen terminal multiplexers / session aggregators and they are all vibed and different, but this one felt so much better to use because of its choices and actually increased the number of parallel sessions I was able to keep going. My own projects I find I have to stress every detail, that is what makes the software good now is really diving into it, the experience feels like just as much of a craft as coding was and 'taste' captures the concept for me.
People who care about their health aren't ignorant about that. I think that can apply to software too: if you care about safety or privacy you might think twice before using vibe-coded software.
I find this post disheartening. To my eye, it seems to be the author's first blog entry with obvious LLM contributions.
Given the recent "moving to a farm" ... 7 days later ... "leaving the farm" posts, it's hard not to read this as someone clearly struggling with the burnout which is all too common in our industry, and resultingly giving in to the false promise of a machine that can just handle the thinking for you.
I keep seeing people trying to define judgement and taste and then they don’t actually like cite Kant’s Critique of Judgement where this is basically the entire concern.
I'm not only referring to that, I'm also referring to a reductionist view when I read "Taste Is All That's Left" - which is a statement that has been mindlessly parroted around as of late.
Yeah even if Dev Schumer is some absurd hype artist he was correct to note that it is judgement and taste, not taste alone, that is most key now. Although he attributed it to the models own subjectivity.
Hard disagree on "Good enough", most things are mediocre at best. Despite the massive AI boom I can at best count on one hand "fresh" products made with AI that are meaningful in day-to-day usage. Not to mention most of them seem to be targeting AI-space itself instead of bridging some actual business use case. Just look at all the products on ShowHN, you can "tell" a vibed product in seconds.
Taste was all that was left when product design peaked in like 2010-12 too and it was a great time.
But the advent of design systems and software like Sketch and Figma pretty much automated the job away. You didn’t need the raw Photoshop skills anymore - any CMU grad could be a top tier designer now.
I see the same thing basically happening here. Taste will matter for a while but eventually people are just choosing the most accepted/expected optimal choices and anything that falls outside the trend will be “not good” in the same way skeuomorphism in design is no longer good.
So we’re lucky we’re in a taste era - it’s a lot of fun. You (yes you!) can still change it, shape it, establish the status quo of tomorrow
Design is not moving pixels around in Photoshop or Figma. CMU grad could be a designer, but to be a great designer you do have to have... well, taste. Figma didn’t change that one bit.
I tend to like images in a split second, when they try to achieve a certain aesthetic at least. I can later inspect them more carefully and understand why they're good.
Reading text, it can either grab me or it doesn't. I've asked about that to an AI and there's some reason for that. Some text is more compatible with me, due to the rhythm or combination of chars/words used, how they actually look/feel, it's not just about the content. So I'm assuming some writers are going to be more compatible with me while others will make me dismiss them quickly or just skim.
I like to think I have the correct objective taste in music, but everybody thinks that about themselves.
yeah the skill du jour keeps shifting. Everyone wants to say they have something the llm doesn't. The more subjective and ill-defined that "something" is the better.
If we define taste as the intuitive act of saying "no, again," then I fully disagree with this whole article.
Every AI-frustrated (but LLM-written, sigh) blog post about the loss of taste and craft and hard work in development sounds like we've given up the interaction with the machine. Like human work is sitting back in your chair and shitting on stuff.
That's obviously not true! That's not how any of this works!
It's hard and weird to develop with the LLMs because they just do stuff. Lots of it is good, some of it is OK, some of it's horrible. Unpacking what it's done is hard and weird because software isn't just lovely UX, it's also data structures that scale and performance and privacy and enterprise controls and SOC 2 and onboarding and accessibility.
If you want to build real software, all that stuff has to get done. Today you're working on the feature, tomorrow you're making it scale. It's long-term and iterative and complex and hard to pack into a prompt or a markdown spec.
The work is the work, done at and with the computer, and it's way more than just "taste."
People will always want something to gatekeep their chosen field. Code went away, so now people need to be able to say if the machine wrote good code. In a year or two people won't be able to use that because the code will be measurably better than something a typical dev would write, and in five years it'll be better than anything anyone could write.
Then we'll need to accept that coding is better when it's done by computers, and we'll start telling everyone it's actually the prompts and context that steered the machine that really matter.
This post is awful! Annoyingly written and zero insight. Most posts about this taste concept are bad (and there have been so many), but this is below average.
> The market timed you both with the same stopwatch and it did not see the difference. It cannot see the difference. Taste does not show up in the diff.
Doesn't that mean guaranteed failure in the market if you use taste as your guiding star? Therefore, there's really no way to enter the market anymore.
I don't know how you then ship software or monetise it...? The market will just clone it.
So... Don't share stuff publicly? Only share with closed groups? Where does this leave small-time software development...?
Someone can replicate any high fashion piece of clothing you make by hand. Maybe they could even make it better. Buyers still care about the origin, just in a pretty small market for it.
Maybe software moves to this after it's commoditized? Users might prefer an artisanal one vs ordering it off the shelf or using an agent themselves.
I keep coming back to this turn. It is not consolation but a correction.
This article is unfortunately full of vague slop like this. It seems the author doesn’t trust their own taste.
I also completely disagree with the premise though - so many seem desperate to discuss the implications of a singularity which shows no sign of arriving.
LLMs have ingested the entire internet and are able to generate plausible and mostly accurate text on demand.
What is missing from thee output of LLMs:
intelligence, agency, motivation, morality, discrimination and yes, taste.
There's a lot here that I enjoy muddling on. And I think taste is very vital. Taste, ambition, seeing possibility: these resonate to me, long resonate, as what I've wanted to see.
And honestly: I've been tired of an age of engineers that are beat down, that only want simple, that finding the most blunt approach is the only way. It has felt tasteless. Anti-ambitious. It often is still very sensible and practical and what you should shoot for! But there are also people wandering around now trying much harder! And I am excited for those frontiers! I think it will challenge and shake the foundations that we've accepted as true, as There Is No Alternative, in exciting bold and fun ways.
I am however pretty unconvinced by the article. There's a lot that doesn't quite work for me, that isn't building the case I'd like, whose takes are off from my read.
> The output is good enough—that is the problem. Good enough is a solvent. It dissolves the reason to do better. For as long as making things was expensive, the expense did quiet work on our behalf. It rationed output.
Agreed about the rationing of output. I do think that the article though continues to show a magical thinking. That we have these things now and they autonomously do the thing. That the LLM's have solved it all.
The article itself goes on at length about how nebulous and abstract taste is:
> Taste is that. It is the compressed, wordless verdict you reach faster than you can justify. It is the “no, again” you say to yourself with total conviction and no available argument
But this contrasts so distinctly to me against "the output is good enough". Is it? That depends. That depends on your taste. The proximate first results come quick. But the technics underneath? Those themselves, in my view, rely enormously on engineering taste to support and advance. I think we see a very similar sort of magical delivery thinking, very clearly on display here:
> When the factories came, they could suddenly make everything—cheaply, uniformly, by the thousand.
As if there was some magical "good enough" transition where suddenly the aliens came and gave us this box that just does the thing. As if we discovered the right formulas and math and now: we had production. Again I think there's just an enormous amount of work and taste that is still actively required inside the factory box, that building the industrial processes is incredibly intense & difficult, even though we have reliable industry-line production and now robots doing the labor.
There's still so many gems, so much lovely material for thought. I love the provocations here, and I think there's a lot of great calibration.
> The friction was not an obstacle to developing taste. The friction was the curriculum.
I agree with your criticism of "the good enough" and "done with taste" dichotomy.
I would like to comment on this, though:
> that only want simple, that finding the most blunt approach is the only way. It has felt tasteless. Anti-ambitious.
I am always aiming for simple. If it's simple, it's maintainable and can be easy to reason about. Doing that is hard work, though. I used to spend three iterations to achieve it: functions are simple, easy to reason about, properly named and composable.
It's the opposite of blunt, though, so it might be that you meant something else with the term "simple" :)
TFA's title is: "Taste Is All That's Left" but when I read the following:
> But the ratio was never the danger. It held steady for centuries. What held the flood back was that producing the crap cost something. Bad novels still took a year to write. Bad software still took a month to build.
I want to rename the title to "Blogs writing in LLMish are all that's left".
Seriously, take this one: "But the ratio was never the danger.".
But understanding the GP was never the danger. Incomprehensible posts on HN have held steady for decades. What held the flood back was that producing crap was all the users could muster. Bad comments took seconds to write. The ten percent was throttled by lack of use of tools like LLMs to help refine the words into something that could be understood.
More importantly, why are there always 30 people who insist that you can’t know it’s claudeslop? Or, “so what if it is,” or something like this? It boggles the mind.
What boggles your mind is not that people are falling for the genetic fallacy, but that others are calling them out for falling for it?
Does this stem from you believing that there is value in making logical errors, or is your angle more towards believing that nobody should waste their time interacting with someone who makes logical errors?
Do you mean starting sentences with "But"? I do, but then again I'm not a native speaker and my writing may be nonstandard. Or do you mean short and direct sentences? In which case, Hemingway.
I'm not saying the article isn't AI slop, but I think it's not obvious from the specific sample you chose, since plenty of people online write like that, before AI.
> the LLMs were trained to mimic how people speak (write)
It's a foundational model (fresh from autoregressive pretraining) that approximates the probability distribution of human texts. And, no, it's not the statistical average of how people speak. It approximates how a person who could have written a text in its context would have written the next words.
Fine-tuning, RLHF, reinforcement learning change this probability distribution. I guess, it's mostly RLHF that shapes the way LLMs write. The similarity of style is due to common providers of RLHF data.
That reminds me of story where the US Air Force attempted to design a cheaper cockpit and chair that would fit the "average pilot".. but in practice it fit nobody well. They went back to the drawing-board and put in all sorts of adjustable components instead.
This is likely because the individual measures being averaged like "femur length" were not independent from one-another, even where they had the benefit of being normally distributed.
The specific quote that we are talking about reads like something I wouldn't find unusual in the works of an individual. Maybe not from someone who stands out as someone who writes things that you'd actually want to read, but from an average Joe? It is likely. I, being no wordsmith myself, have written things just like that. It looks quite typical of how people write to me.
In fact, I find no meaningful difference between "But no _individual_ speaks like this." and "But the ratio was never the danger.", aside from them being about different topics. They are syntactically very similar.
The person you're replying to is saying that it wasn't written. If you are the author, I will believe you that it was. If you are not the author, I have no idea why you think you know more than anyone else does.
Even if they didn't make a fair point (and I think they do), I would prefer their pedantry to your palpable contempt by a large margin. I mean "palpable" literally. My body is tenser and my heart more sad having experienced the voice you just used on that person.
These 100% AI generated posts about taste have to be some elaborate performance art piece, I genuinely do not understand how someone could think it a good idea "write" this.
In either case, taste being all that's left doesn't really help at all since whatever FAANG or top startups do is by definition good taste. Same for fashion, art and so on with the top being suitably defined. Slim fit was all the rage until all the uncool people got in and that's what ended the trend and not any friction or utility.
It's also just patently false. Taste is not all that is left. Are software engineers really so full of hubris that they thought coding is all that there is to making a product?
> Are software engineers really so full of hubris that they thought coding is all that there is to making a product?
Frankly, this barely scratches the surface of the hubris of engineers. I personally think AI has been one of the best thing to happen to software engineering. Writing boilerplate or my millionth auth implementation was never why I fell in love with the craft in the first place.
We have computers, computers can automate things. Boilerplate is a sign of a bad language (not a bad engineer). But then we add infinite automation features to reduce boilerplate and we get C++, and C++ programmers decry the excessive number of features and want to delete them all and go back to writing boilerplate by hand.
I once wrote a pile of C++ templates that would let you wrap a C++ function as a JavaScript function, automatically converting all the arguments and the return type. Then I left the company. How did my successors find it? No idea. Maybe they pulled out all the automation complexity and replaced it with manually operated boilerplate.
[delayed]
> Boilerplate is a sign of a bad language
This is true, but I was mostly referring to framework boilerplate (Spring, React, etc.) or plumbing boilerplate. If you've ever written code professionally, you know that most code that you write is just making your thing fit with someone else's thing.
If your consumer or your provider made bad engineering decisions, you have to absorb them. If your company mandates the use of a framework, you have to absorb that, and so on. So I think it's great that AI can write all the dumb shit I didn't come up with anyway.
Spewing out tons of boilerplate has always been the hallmark of a bad engineer.
What is left? The second-best architecture for any given technical problem is, in the majority of cases, still good enough.
When all else breaks, absurd() is your friend.
So i guess proven secure architecture will come and stay.
I think I am over opinion articles on AI, as a genre, in general.
Are you saying they are... tasteless?
> So let me be plain about it: Claude was not here. No LLM wrote this—not a sentence of it, nor was it outlined, drafted, reviewed, checked, etc. by one, and there is no prompt behind it either
Following form of the words I just read: these 100% animous filled HN threads about how everyone is bad and everything deserves mocker have to be some elaborate power game to undermine hope and human will, I genuinely do not understand how genuine human people could think it good to "comment" like this.
There's just such a ridiculous pile on of people with such enormous overwhelming negativity underneath this ocmment. This off-gassing is almost never engaged with the material. Now there's an excuse to write off any material we don't like instantly, to accuse it of fakeness. But I still do not see a realness enscribed upon people who so actively work to tear down & prevent thought & consideration.
To engage with the post: I do think the definition of fashion is interesting, and a real power dynamic to be aware of. But do think there's a lot of things that take off in the world because they are good and interesting and nice, especially in technology and products, that the world wants nice things, and will happily support you for doing that. And especially in tech, the technical decisions we start from and build underfoot steer us towards success in subtle compounding ways, that matter, and that taste is crucial, where fashion is not a factor.
Again, at a meta level: my distress & frustration at the top is less from seeing this comment, and far more from seeing everything down thread trying to out-do one another with how cynical and anti-meaning anti-trying they can be. But I think this post definitely started on an all too familiar note, blood in the HN waters that all too often gets us so sharky.
IMO giving "benefit of the doubt" to something beyond doubt is bad, straight up. If someone can't tell this article is AI then they need to go spend an hour a day reading known AI writing vs known human writing until they develop the skill; the voice of this author SCREAMS AI.
I'm a firm believer in not engaging with or promoting the content of misrepresented AI.
On the topic of taste, I think of this quote in Susan Sontag's "Notes On 'Camp'" often:
“For taste governs every free — as opposed to rote — human response. Nothing is more decisive. There is taste in people, visual taste, taste in emotion — and there is taste in acts, taste in morality. Intelligence, as well, is really a kind of taste: taste in ideas. (One of the facts to be reckoned with is that taste tends to develop very unevenly. It's rare that the same person has good visual taste and good taste in people and taste in ideas.)
Taste has no system and no proofs. But there is something like a logic of taste: the consistent sensibility which underlies and gives rise to a certain taste. A sensibility is almost, but not quite, ineffable. Any sensibility which can be crammed into the mold of a system, or handled with the rough tools of proof, is no longer a sensibility at all. It has hardened into an idea..."
This article resonated tremendously with me, speaking as someone who has been coding since the 1980s and had to develop my own taste the hard way, through lots and lots of mistakes. Whenever I see a software demo built by someone mostly using agents to do the work, I have to question whether it's any good on the inside - or whether any real intuition or judgment went into its design. On the flip side, maybe none of that matters anymore - as long as it works, who cares how it was built.
It resonated with me too, and even as a once fairly prolific coder that now uses agents to do all the coding, I feel like this is all relevant and does matter. The agent does whatever you ask and provides thin judgements based on what it thinks is your motivation and context, even feeding it the context it will often make bad generic choices. I think this is the part hard to optimize because it's not a verifiable reward, it's the feel of the software. I think Herdr is a great example, I've tried several dozen terminal multiplexers / session aggregators and they are all vibed and different, but this one felt so much better to use because of its choices and actually increased the number of parallel sessions I was able to keep going. My own projects I find I have to stress every detail, that is what makes the software good now is really diving into it, the experience feels like just as much of a craft as coding was and 'taste' captures the concept for me.
I wonder, when you have a McDonald's McDouble, do you really care how it was made or do you just care that it's yummeroni and costs $2.99?
People who care about their health aren't ignorant about that. I think that can apply to software too: if you care about safety or privacy you might think twice before using vibe-coded software.
I find this post disheartening. To my eye, it seems to be the author's first blog entry with obvious LLM contributions.
Given the recent "moving to a farm" ... 7 days later ... "leaving the farm" posts, it's hard not to read this as someone clearly struggling with the burnout which is all too common in our industry, and resultingly giving in to the false promise of a machine that can just handle the thinking for you.
I keep seeing people trying to define judgement and taste and then they don’t actually like cite Kant’s Critique of Judgement where this is basically the entire concern.
Speaks more to the myopia of developers than anything else.
Most people in Comp Sci don’t read Kant unless they have a concentration in Artificial Intelligence, so I’m not blaming them.
I'm not only referring to that, I'm also referring to a reductionist view when I read "Taste Is All That's Left" - which is a statement that has been mindlessly parroted around as of late.
Yeah even if Dev Schumer is some absurd hype artist he was correct to note that it is judgement and taste, not taste alone, that is most key now. Although he attributed it to the models own subjectivity.
Hard disagree on "Good enough", most things are mediocre at best. Despite the massive AI boom I can at best count on one hand "fresh" products made with AI that are meaningful in day-to-day usage. Not to mention most of them seem to be targeting AI-space itself instead of bridging some actual business use case. Just look at all the products on ShowHN, you can "tell" a vibed product in seconds.
Mediocre is good enough, that's what it means.
Mediocre literally means middling, which may or may not be good enough depending on requirements.
Taste was all that was left when product design peaked in like 2010-12 too and it was a great time.
But the advent of design systems and software like Sketch and Figma pretty much automated the job away. You didn’t need the raw Photoshop skills anymore - any CMU grad could be a top tier designer now.
I see the same thing basically happening here. Taste will matter for a while but eventually people are just choosing the most accepted/expected optimal choices and anything that falls outside the trend will be “not good” in the same way skeuomorphism in design is no longer good.
So we’re lucky we’re in a taste era - it’s a lot of fun. You (yes you!) can still change it, shape it, establish the status quo of tomorrow
Design is not moving pixels around in Photoshop or Figma. CMU grad could be a designer, but to be a great designer you do have to have... well, taste. Figma didn’t change that one bit.
Going to drop this https://commoncog.com/the-tacit-knowledge-series/. Tacit experience gives you taste in my opinion.
This post sucks, and it was also written by Claude.
I tend to like images in a split second, when they try to achieve a certain aesthetic at least. I can later inspect them more carefully and understand why they're good.
Reading text, it can either grab me or it doesn't. I've asked about that to an AI and there's some reason for that. Some text is more compatible with me, due to the rhythm or combination of chars/words used, how they actually look/feel, it's not just about the content. So I'm assuming some writers are going to be more compatible with me while others will make me dismiss them quickly or just skim.
I like to think I have the correct objective taste in music, but everybody thinks that about themselves.
Everybody thinks they have good taste don’t they.
How convenient it is that we’re all so awesome according to the one metric that can’t be measured
yeah the skill du jour keeps shifting. Everyone wants to say they have something the llm doesn't. The more subjective and ill-defined that "something" is the better.
> Everyone wants to say they have something the llm doesn't.
LLM can't waste weeks debating stacks and tools and then recast the problem as a new framework that needs to exist.
*Check mate*
Very well written, this gave me goosebumps. Full of flavor.
No, the AI is writing code not architecture. And it's the same for all disciplines.
There will be no mass job layoffs for antying but relatively automatable stuff.
If we define taste as the intuitive act of saying "no, again," then I fully disagree with this whole article.
Every AI-frustrated (but LLM-written, sigh) blog post about the loss of taste and craft and hard work in development sounds like we've given up the interaction with the machine. Like human work is sitting back in your chair and shitting on stuff.
That's obviously not true! That's not how any of this works!
It's hard and weird to develop with the LLMs because they just do stuff. Lots of it is good, some of it is OK, some of it's horrible. Unpacking what it's done is hard and weird because software isn't just lovely UX, it's also data structures that scale and performance and privacy and enterprise controls and SOC 2 and onboarding and accessibility.
If you want to build real software, all that stuff has to get done. Today you're working on the feature, tomorrow you're making it scale. It's long-term and iterative and complex and hard to pack into a prompt or a markdown spec.
The work is the work, done at and with the computer, and it's way more than just "taste."
People will always want something to gatekeep their chosen field. Code went away, so now people need to be able to say if the machine wrote good code. In a year or two people won't be able to use that because the code will be measurably better than something a typical dev would write, and in five years it'll be better than anything anyone could write.
Then we'll need to accept that coding is better when it's done by computers, and we'll start telling everyone it's actually the prompts and context that steered the machine that really matter.
In couple of years data compression will have progressed so much that you ll be able to store all the data in the world on a floppy disk!
This post is awful! Annoyingly written and zero insight. Most posts about this taste concept are bad (and there have been so many), but this is below average.
> The market timed you both with the same stopwatch and it did not see the difference. It cannot see the difference. Taste does not show up in the diff.
Doesn't that mean guaranteed failure in the market if you use taste as your guiding star? Therefore, there's really no way to enter the market anymore.
I don't know how you then ship software or monetise it...? The market will just clone it.
So... Don't share stuff publicly? Only share with closed groups? Where does this leave small-time software development...?
Someone can replicate any high fashion piece of clothing you make by hand. Maybe they could even make it better. Buyers still care about the origin, just in a pretty small market for it.
Maybe software moves to this after it's commoditized? Users might prefer an artisanal one vs ordering it off the shelf or using an agent themselves.
Well, this has been true (to a lesser extent) for decades. See "Worse Is Better" and all the subsequent essays and writing following on from that
I keep coming back to this turn. It is not consolation but a correction.
This article is unfortunately full of vague slop like this. It seems the author doesn’t trust their own taste.
I also completely disagree with the premise though - so many seem desperate to discuss the implications of a singularity which shows no sign of arriving.
LLMs have ingested the entire internet and are able to generate plausible and mostly accurate text on demand.
What is missing from thee output of LLMs: intelligence, agency, motivation, morality, discrimination and yes, taste.
I don't think any of these tech people talking about 'taste' have any. Least of not a person with an anime profile image in their about me.
Intentional anonymity is perfectly tasteful.
Taste good, taste hard to cultivate, taste easy to fake and hard to prove. There, TLDR.
We still have a metric for good taste, and it is $. Solve a problem that people care about? They give you $. Solve problem no one care about? No $.
It is not perfect, but I challenge you to find a better metric that is as effective and simple.
There's a lot here that I enjoy muddling on. And I think taste is very vital. Taste, ambition, seeing possibility: these resonate to me, long resonate, as what I've wanted to see.
And honestly: I've been tired of an age of engineers that are beat down, that only want simple, that finding the most blunt approach is the only way. It has felt tasteless. Anti-ambitious. It often is still very sensible and practical and what you should shoot for! But there are also people wandering around now trying much harder! And I am excited for those frontiers! I think it will challenge and shake the foundations that we've accepted as true, as There Is No Alternative, in exciting bold and fun ways.
I am however pretty unconvinced by the article. There's a lot that doesn't quite work for me, that isn't building the case I'd like, whose takes are off from my read.
> The output is good enough—that is the problem. Good enough is a solvent. It dissolves the reason to do better. For as long as making things was expensive, the expense did quiet work on our behalf. It rationed output.
Agreed about the rationing of output. I do think that the article though continues to show a magical thinking. That we have these things now and they autonomously do the thing. That the LLM's have solved it all.
The article itself goes on at length about how nebulous and abstract taste is:
> Taste is that. It is the compressed, wordless verdict you reach faster than you can justify. It is the “no, again” you say to yourself with total conviction and no available argument
But this contrasts so distinctly to me against "the output is good enough". Is it? That depends. That depends on your taste. The proximate first results come quick. But the technics underneath? Those themselves, in my view, rely enormously on engineering taste to support and advance. I think we see a very similar sort of magical delivery thinking, very clearly on display here:
> When the factories came, they could suddenly make everything—cheaply, uniformly, by the thousand.
As if there was some magical "good enough" transition where suddenly the aliens came and gave us this box that just does the thing. As if we discovered the right formulas and math and now: we had production. Again I think there's just an enormous amount of work and taste that is still actively required inside the factory box, that building the industrial processes is incredibly intense & difficult, even though we have reliable industry-line production and now robots doing the labor.
There's still so many gems, so much lovely material for thought. I love the provocations here, and I think there's a lot of great calibration.
> The friction was not an obstacle to developing taste. The friction was the curriculum.
Makes me think of yesterdays @apenwarr banger,
> Every slow prototype started out as a fast prototype, I think that’s how it goes https://bsky.app/profile/apenwarr.ca/post/3msemlo4rds2h
I agree with your criticism of "the good enough" and "done with taste" dichotomy.
I would like to comment on this, though:
> that only want simple, that finding the most blunt approach is the only way. It has felt tasteless. Anti-ambitious.
I am always aiming for simple. If it's simple, it's maintainable and can be easy to reason about. Doing that is hard work, though. I used to spend three iterations to achieve it: functions are simple, easy to reason about, properly named and composable.
It's the opposite of blunt, though, so it might be that you meant something else with the term "simple" :)
The Rick Rubin meme
would you let taste make withdrawls from your bank account
i type code by hand
so emacs is also left :)
My non-dominant hand is also left.
AI slop
TFA's title is: "Taste Is All That's Left" but when I read the following:
> But the ratio was never the danger. It held steady for centuries. What held the flood back was that producing the crap cost something. Bad novels still took a year to write. Bad software still took a month to build.
I want to rename the title to "Blogs writing in LLMish are all that's left".
Seriously, take this one: "But the ratio was never the danger.".
Who the FUCK speaks like that?
Hemingway wrote the same way. Here's a famous example from the Old Man and the Sea:
Every day is a new day. It is better to be lucky. But I would rather be exact. Then when luck comes you are ready.
The reason is simple: Short sentences are punchy. They hit hard, carry emotional weight.
P.S. As evident, clearly I'm having some fun with this, but the point stands.
Short sentences were never the danger, or what the GP is referring to.
But understanding the GP was never the danger. Incomprehensible posts on HN have held steady for decades. What held the flood back was that producing crap was all the users could muster. Bad comments took seconds to write. The ten percent was throttled by lack of use of tools like LLMs to help refine the words into something that could be understood.
Correction:
> The reason is simple. Short sentences are punchy. They hit hard. But they carry emotional weight.
More importantly, why are there always 30 people who insist that you can’t know it’s claudeslop? Or, “so what if it is,” or something like this? It boggles the mind.
What boggles your mind is not that people are falling for the genetic fallacy, but that others are calling them out for falling for it?
Does this stem from you believing that there is value in making logical errors, or is your angle more towards believing that nobody should waste their time interacting with someone who makes logical errors?
Yep, it is slop. I got 2-3 paragraphs in and closed the tab, as too many LLMisms raised their ugly heads.
> Who the FUCK speaks like that?
Do you mean starting sentences with "But"? I do, but then again I'm not a native speaker and my writing may be nonstandard. Or do you mean short and direct sentences? In which case, Hemingway.
I'm not saying the article isn't AI slop, but I think it's not obvious from the specific sample you chose, since plenty of people online write like that, before AI.
> Who the FUCK speaks like that?
If we assume that it is LLM generated, as suggested, apparently a lot of people, as the LLMs were trained to mimic how people speak (write).
You might find it to be unusual because you probably have never read anything written by most people. Only peculiar characters tend to get noticed.
> the LLMs were trained to mimic how people speak (write)
It's a foundational model (fresh from autoregressive pretraining) that approximates the probability distribution of human texts. And, no, it's not the statistical average of how people speak. It approximates how a person who could have written a text in its context would have written the next words.
Fine-tuning, RLHF, reinforcement learning change this probability distribution. I guess, it's mostly RLHF that shapes the way LLMs write. The similarity of style is due to common providers of RLHF data.
But no _individual_ speaks like this. It is not common that a guy representative of the Average Of All People Plus Some RL writes a blog post.
That reminds me of story where the US Air Force attempted to design a cheaper cockpit and chair that would fit the "average pilot".. but in practice it fit nobody well. They went back to the drawing-board and put in all sorts of adjustable components instead.
This is likely because the individual measures being averaged like "femur length" were not independent from one-another, even where they had the benefit of being normally distributed.
The specific quote that we are talking about reads like something I wouldn't find unusual in the works of an individual. Maybe not from someone who stands out as someone who writes things that you'd actually want to read, but from an average Joe? It is likely. I, being no wordsmith myself, have written things just like that. It looks quite typical of how people write to me.
In fact, I find no meaningful difference between "But no _individual_ speaks like this." and "But the ratio was never the danger.", aside from them being about different topics. They are syntactically very similar.
> Who the FUCK speaks like that?
Walter Hartwell White, I suppose.
> Who the FUCK speaks like that?
It's a blog post. It was written. The written word is different from the spoken.
People don't speak like: The man in black fled across the desert, and the gunslinger followed.
People don't speak like: In a hole in the ground there lived a hobbit.
In fact, I can throw this question at you:
"TFA's title is"
Who the FUCK speaks like that? Who the FUCK uses TFA's IRL OMG WTF?
Silly, isn't it?
The sky above the port was the color of television, tuned to a dead channel.
> It's a blog post. It was written.
The person you're replying to is saying that it wasn't written. If you are the author, I will believe you that it was. If you are not the author, I have no idea why you think you know more than anyone else does.
Yes, your pedantry shows how you're very clever. Please be very clever somewhere else.
Even if they didn't make a fair point (and I think they do), I would prefer their pedantry to your palpable contempt by a large margin. I mean "palpable" literally. My body is tenser and my heart more sad having experienced the voice you just used on that person.
Thank you, I take great pride and effort in compressing as much emotional weight into as short a statement as possible.
I miss when people used to hide emotional fragility, and not publicly brag about being physically and emotionally devastated by an HN comment.
Please be nostalgic somewhere else.